Yueh Chen Chiang¹ (), Yared R. Haile¹, Christopher R. Fetsch¹; ¹Zanvyl Krieger Mind/Brain Institute & Solomon H. Snyder Department of Neuroscience, Johns Hopkins University

In natural environments, animals must integrate information from multiple modalities to guide action, while also balancing decision speed vs. accuracy. Although the neural basis of multisensory integration has been extensively studied, previous work mainly used fixed-duration tasks which limit the ability to isolate the temporal window of decision formation on a trial by trial basis. We trained rhesus monkeys to perform a reaction-time (RT) heading discrimination task based on vestibular cues (translational inertial motion), visual cues (optic flow), or both. Animals reported a binary heading judgment with a saccadic eye movement as soon as they were ready, followed by a binary confidence rating with a second saccade (post-decision wagering). Neural recordings were obtained from multisensory cortical areas MSTd and VPS using 32-channel linear arrays (N=440 units in 2 monkeys) or Neuropixels probes (N=932 units in 1 monkey). At the population level, both areas showed clear condition-dependent structure, with stimulus, choice, and confidence components separable in low dimensional space. However, single-unit stimulus encoding was notably weak in the RT task. Within the time frame of individual decisions, neurometric thresholds averaged between 55 and 90 degrees, 2.5-fold higher than during a passive fixation task (after controlling for stimulus duration) and 18-fold higher than psychophysical thresholds. Population decoding confirmed the relatively poor fidelity of heading signals in the RT context, and revealed that choice and confidence signals emerged only in the final few hundred milliseconds before the first saccade. The results suggest that either (1) momentary evidence for speeded heading decisions resides elsewhere in the brain, or (2) perceptual readout is targeted to a small subset of neurons and/or a brief window before commitment to the choice. We consider these results in the context of alternatives to the standard evidence accumulation framework implemented within a circuit model of the visual-vestibular cortical network.

Acknowledgements: Supported by NIH-NINDS (RF1NS132910), the E. Matilda Ziegler Foundation for the Blind, the Whitehall Foundation, the France-Merrick Foundation